‘Very deep flaws’ in generative AI tech, says boss of ChatGPT firm OpenAI

18 January 2024, 16:04

Sam Altman
PA-74419769 Cropped (1). Picture: PA

Sam Altman added that, despite rapid improvements in generative AI, ‘you don’t actually want that to drive your car’.

There are “very deep flaws” in current generative AI technology and the public is learning its limitations, according to the boss of ChatGPT’s parent business.

Sam Altman, chief executive officer of OpenAI, added that, despite rapid improvements in generative AI, “you don’t actually want that to drive your car”.

The comments came as he answered questions on a panel alongside Chancellor Jeremy Hunt, Salesforce chief Marc Benioff, Accenture chief Julie Sweet and Pfizer chief Albert Bourla at the World Economic Forum in Davos.

Mr Altman told the audience that people are becoming accustomed to using AI tools and that they “know it might be wrong but that can be ok”.

“It’s a very limited current capability and it has very deep flaws,” he said.

“People are finding ways to use it for great productivity gains and other gains, and they understand the limitations.

“A system that is sometimes right, sometimes creative but often totally wrong – you don’t actually want that to drive your car.

“But you are happy for it to help you brainstorm what to write about or help you with code that you then check.”

“The OpenAI style of model is good at some things but not good for a life-and-death situation.”

It comes just two months after the co-founder and chief executive officer dramatically returned to the helm of the business after OpenAI’s board of directors had voted no confidence in him days earlier.

Mr Altman admitted the company’s board was too small and did not have the necessary experience.

“At some point you just have to laugh, at some point, it just gets so ridiculous,” he said, describing the period around his reinstatement.

By Press Association

More Technology News

See more More Technology News

National Cyber Security Centre launch

National Cyber Security Centre names Richard Horne as new chief executive

The lights on the front panel of a broadband internet router, London.

Virgin Media remains most complained about broadband and landline provider

A person using a laptop

£14,000 being lost to investment scams on average, says Barclays

Europe Digital Rules

Meta unveils latest AI model as chatbot competition intensifies

AI technology

Younger children increasingly online and unsupervised, Ofcom says

Migrant Channel crossing incidents

Ministers will be told to use AI to screen migrants for threats, adviser says

Nothing smartphone

UK tech firm Nothing to integrate ChatGPT into its devices

The Google offices in Six Pancras Square, London

Google confirms more job cuts as part of company reorganisation

Person using laptop

Housing association reprimanded after residents’ data compromised

A screengrab of an arrest in connection with the LabHost website

Arrests made and thousands of victims contacted after scammer site taken offline

Social media apps on a smartphone

Three-quarters of public fear misinformation will affect UK elections – report

Businessman racing with a robot

TUC calls for AI to be regulated in the workplace

The ChatGPT website

AI chatbot ‘could be better at assessing eye problems than medics’

FastRig wingsail launch

Scottish-made wingsail set for sea tests after launch on land

Immigration

Rollout of eVisas begins as Government aims for digital immigration by 2025

Elon Musk in 2024

X may start charging new users to post, says Elon Musk